I have not been able to come up with a way to overcome p!=np in logic, so I am thinking about developing a heavily typed logic (logic-like references) as a way to get around the bottleneck of exponential complexity. However, I have run into some difficulties there as well.
I would like an object (a concept or a concept fragment) to be able to inherit from different type categories. The categorical types would be familial categories meaning that every object (of that type) possessed some features of the type family but they would not necessarily possess all the features of the family. Subtypes (composed of collections of objects) would be used to better represent the type hierarchies. Relations between objects or between types would be treated as conceptual objects themselves. The fluidity of this kind of typing would make the system more natural than old AI but it would be much more difficult to program. I initially want to use this system as a way of cross-indexing a text document that a user could set up. It is my hope that I would be able to develop an AI system that would be capable of learning to read – to some extent – based on the conceptual relations that the user taught it (and continued to teach it). One idea that I had is that I might use non-familial categorical types right at the level above the conceptual objects. The objects of the non-familial category type would inherit (or possess) all the features of the type. Then if a set of features were triggered in an individual (concept object) the sub-family types might be identified and in a reverse manner, if the sub-family typing were pointedly used as representations of features then a better assessment of the conceptual objects that possessed those features (or of the type of objects that would possess those objects) could be made. This would be useful in learned references. I am thinking about working on an easily feasible program that would actually be useful. Then once I got into it I might be able to develop the idea to give it a genuine AI feature. I think text based search engine look-up has actually deteriorated a little during the past 10 years. I just looked up familial categories and my first reference was https://en.wikipedia.org/wiki/Category:Family . Going with categorical family I came up with https://en.wikipedia.org/wiki/Categorical_distribution and https://plato.stanford.edu/entries/category-theory , none of which are actually related to what I am talking about. I finally put, “categorical family Wittgenstein” and I found one reference on Stack Exchange that looks like it might be a little relevant. https://philosophy.stackexchange.com/questions/40581/on-wittgensteins-family-resemblance-and-machine-learning ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T8c37d01ebc4fffa8-M1dc409ffb726f4ba8a216208 Delivery options: https://agi.topicbox.com/groups/agi/subscription
